USE OF REMOTE SENSING AND GIS FOR FLOOD HAZARD
MAPPING IN CHIANG MAI PROVINCE, NORTHERN THAILAND
Minyan DUAN a, *, Jixian ZHANG a, Zhengjun LIU a, Aekkapol Aekakkararungroj b
a
b
Chinese Academy of Surveying and Mapping, Beijing 100039, China
Bureau of Research Development and Hydrology, Department of Water Resources, BKK 10400, Thailand
KEY WORDS: flood disaster, hydrologic model, HEC-RAS
ABSTRACT:
Throughout the human history floods have been an integral part of the civilization. Still men have not quite coped well to live with
floods. Flood hazards result from a combination of physical exposure and human vulnerability to a geophysical process. Physical
exposure reflects the type of flood events that can occur and their statistical pattern at a particular site while human vulnerability
reflects key socioeconomic factors such as the number of people at risk, the extent of any flood protection works and the ability of
the population to anticipate and cope with the hazard. Recently, the advancement in computer-aided technology has been extensively
used in formulating models used for flood calculation and hazard analysis. This study focuses on using a hydraulic model HEC-RAS
in a GIS and Remote Sensing environment for the area of Ping River basin, northern Thailand, generates the inundation area and the
flood depth for the year 2005 flood event in Chiang Mai province by using the 1D HEC-RAS flood model, verifies the model by
comparing the model results with the remote sensing image, and prepares hazard maps using the model output and other
socio-economic data.
1.
INTRODUCTION
results of the flood phenomenon and do vulnerability or risk
analysis and hazard assessment in a 1D, 2D, or 3D approach.
Remote sensing can be used to validate the model results by
comparing the flood inundation area. Remote sensing images
are used as base maps for flooded areas since they can be
acquired in all weather condition and at any time. This study
focuses on using a hydraulic model HEC-RAS in a GIS and
Remote Sensing environment for the area of Ping River basin,
northern Thailand.
Flood disasters are among the world’s most frequent and
damaging types of disaster (Parker, 2002). They have been the
most common type of geophysical disasters in the late half of
the twentieth century, generating an estimated more than 30
percent of all disasters between 1945 and 1986 (Glickman et al,
1992; Shah, 1983; Dworkin, 1976; Sheehan and Hewitt, 1973).
The data produced by Glickman et al. (1992) indicate that
globally, flood disasters are about the third most harmful form
of geophysical disaster when the number of deaths concern.
The majority of floods are harmful to human settlements and
yearly flooding, on average, may victimize 20,000 lives and
affect 75 million people (Coburn 1994). Throughout the human
history floods have been an integral part of the civilization. Still
men have not quite coped well to live with floods. This can be
attributed to the complex nature of the flood as well as the
diverse response to it. It is always hard to tell which one is the
better policy and strategy, to fight floods or to learn to live with
the floods. Flood hazards result from a combination of physical
exposure and human vulnerability to a geophysical process.
Physical exposure reflects the type of flood events that can
occur and their statistical pattern at a particular site while
human vulnerability reflects key socioeconomic factors such as
the number of people at risk, the extent of any flood protection
works and the ability of the population to anticipate and cope
with the hazard (Smith and Ward, 1998).
Seasonal flooding is a regular feature of the Monsoon climate
and flood plain landscapes of Thailand. Most of the major cities
in Thailand, including historical and current capitals of
Kingdoms, such as Chiang Mai, Ayutthaya and Bangkok, have
been built on the foundations of rice-growing civilizations in
major flood plains. Communities where lives depend on a
seasonal cycles of flood have learnt to live with floods and
embrace its arrival with songs and dances. Institutions and
cultural practices relating to the management of floods are
persistent and have survived for centuries up to now.
2.
HYDRAULIC MODEL
Hydraulic models are concerned with the dynamics of flow in
channels and in overbank areas. They predict water levels and
velocities in time and space. They use boundary conditions
such as the results of hydrologic models and recorded flood
data (CRC, 2006).
Recently, the advancement in computer-aided technology has
been extensively used in formulating models used for flood
calculation and hazard analysis. This requires mainly two parts.
A hydrologic model to calculate the runoff from the rainfall
and a hydraulic model to determine the water surface profiles at
specific locations along the stream. GIS is used to visualize the
2.1
Types of Hydraulic Models
Hydraulic models can be divided into 1D, 2D quasi 2D or 3D
based on their ability to model in 1D, 2D, or 3D space. 1D
* Corresponding author. duanmy@casm.ac.cn
203
model takes cross sections to describe the geometry of the
channel and streamlines of flow are considered into intersect
the cross sections at right angles. In reality this is not correct, as
there is velocity both in parallel as well as perpendicular to the
direction of the stream. 1D model can be used to represent 2D
behavior by splitting up the 1D cross-section into multiple 1D
cross-section. This is often described as quasi 2D modeling. 2D
models are those that consider both x and y components of the
velocity of flow. This is a closer approximation of the reality as
compared to a1D model. 2D models are well suited where there
are a broad estuaries or wide floodplains. Generally more data
and more computational power is required to carry out a 2D
model simulation. 3D models are those where all the three
components of velocity in x, y and z directions of flow are
considered. But for flood modeling, 3D modeling is not
necessarily considered (USACE. 2002).
component of HEC-RAS simulates one-dimensional unsteady
flow through a full network of open channels and is primarily
used for subcritical flow regime calculations but may also be
applied for supercritical and rapidly varied flows. The unsteady
flow module has some additional capabilities like it can model
storage area and hydraulic connections between storage areas.
The unsteady flow analysis is performed by applying the full
equations of motion (St. Venant equations) at a cross-section
with upstream and downstream boundary conditions and
various other parameters. Data may be directly fed to
HEC-RAS or can be imported from excel spreadsheet. It has a
GUI to facilitate the input and manipulation of data. This
includes a steady flow editor, an unsteady flow editor and a
geometric data editor including a junction editor, a
cross-section editor and various structures and hydraulic
features editors. ArcView GIS also has an extension called
HEC-Geo RAS which is actually an interface between
ArcView GIS and HEC-RAS. It is specifically designed to
process geospatial data for the use with HEC-RAS. No other
hydrologic or hydraulic modeling software results can be
automatically input to HEC-RAS. The geometric data required
to define in HEC-RAS includes:
¾
Cross-section data
¾
Reach lengths (measured between cross sections)
¾
Stream junction information (Reach lengths across
junctions and tributary angles)
UP STREAM
For an unsteady flow, hydraulic structures are modeled by
taking into consideration the physical parameters of the
structure in the appropriate standard structure format in the
HEC-RAS data editor. The types of structures that can be
modeled in HEC-RAS include bridges, culverts, inline and
lateral weirs and gates, spillways and levees. Moreover the
unsteady component can model storage areas, hydraulic
connections between storage areas, hydraulic connections
between stream reaches, pumping stations, flap-gated culverts.
Other features include calculation of ineffective areas,
floodplain encroachment analysis, channel modification
analysis, scour analysis at bridges, dam breaching and levee
breaking algorithm, groundwater interflow and contraction and
expansion losses. Post processing capabilities include:
¾
Longitudinal profiles: The user can view the water
surface profiles along the length of the channel for each
flow profile.
¾
Profile Plots: The user can view the profiles of various
parameters such as velocity, flow and depth against
longitudinal chainage.
¾
Rating curves: The user may view the computed rating
curves at each cross-section.
¾
Perspective Plot: The user may view a 3D perspective
view of the river system and the water surface profiles.
¾
Flow and stage hydrographs: The user may visualize
flow and stage hydrographs at each cross section for
unsteady flow simulation.
¾
Output tables: Detailed and summarized output tables
of various parameters may be viewed and exported.
Figure 1. 1D (left) and 2D (right) representations of open
channel flow
Another two different types of models are the finite difference
model and finite element model. The finite difference model
uses a network consisting of a regular grid of nodes to represent
the catchment information and the finite element model uses a
network that consists of an irregular mesh of nodes where the
elements consists of triangles and quadrilaterals.
The major advantage of the finite difference method is that it is
relatively simple to set up and operate. The finite element
method is better able to handle complex geometries and
boundaries while the finite difference method is restricted to
handle rectangular shapes and simple modifications of such
shapes (USACE. 2002).
2.2
Overview of HEC-RAS
The Hydrologic Engineering Centre’s River Analysis System,
(HEC-RAS) was developed by the U.S. Army Corps of
Engineers (USACE) led by Gary W. Brunner. It was released
in 1995 and it was a replacement for HEC-2 which was an
earlier version and widely used since 1970. From 1995 to 2000,
HEC-RAS could only be used for steady, gradually varied flow
modeling. The capability of unsteady flow modeling was added
in 2001. After several modifications the current version of
HEC-RAS is 3.1.3 and it is available for free download from
the HEC-RAS website (Review and Assessment of
Hydrologic/Hydraulic Flood Models, 2006).
With the help of HEC-GeoRAS, the user can export the
HEC-RAS data to Arc-View for performing more calculations
such as flood inundation and hazard mapping (Review and
Assessment of Hydrologic/Hydraulic Flood Models, 2006).
HEC-RAS model can perform water surface calculations for
gradually varied steady flow for a river reach, a dendritic
system, or a full network of channels. This steady flow
component is capable of modeling subcritical, supercritical and
mixed flow regime water surface profiles. The steady flow
component is based on the solution of the 1D energy equations.
A peak discharge is applied at each cross section to determine
the maximum water surface elevation. The unsteady flow
2.3
Analytic Hierarchy Process (AHP)
The Analytical Hierarchical Process (AHP) is a multi-criteria
decision making technique, which provides a systematic
approach for assessing and integrating the impacts of various
204
factors, involving several levels of dependent or independent,
qualitative as well as quantitative information.
The weather in Chiang Mai is relatively cool all year round,
with an average temperature of 25ºC. Temperatures typically
range between 20ºC and 31ºC. The relative humidity averages
72%, and annual rainfall is normally 1,000-1,200 mm.
It is a methodology to systematically evaluate, often conflicting,
qualitative criteria (Saaty, 1980). Like other multi-attribute
decision models, AHP also attempts to resolve conflicts and
analyze judgments through a process of determining the
relative importance of a set of activities or criteria by pairwise
comparison of these criteria on a 9-point scale. In order to do
this, a complex problem is first divided into a number of
simpler problems in the form of a decision hierarchy (Erkut and
Moran, 1991). AHP is often used to compare the relative
preferences of a small number of alternatives concerning an
overall goal. AHP is becoming popular in decision-making
studies where conflicting objectives are involved. Recently,
Siddiqui et al., (1996) introduced a new method known as
Spatial – AHP to identify and rank areas that are suitable for a
landfill, using knowledge based user preferences and data
contained in GIS maps.
3.2
Infrequent large floods usually occur in northern Thailand late
in the May–October rainy season. Although the May–October
rainfall is dominated by storms of moist air moving northeast
from the Indian Ocean, large floods are typically associated
with tropical 25 depressions moving westward from the South
China Sea.
Date
13-16 Aug.
3.
3.1
Floods in Chiang Mai
20-22 Sept.
STUDY AREA
29 Sept. –
1 Oct.
Background
Description
A heavy monsoon rainstorm associated
with a low-pressure trough moving
westward across northern Thailand
Tropical storm Vincente weakened to a
tropical depression traveling westward
across Indochina from the South China Sea
Typhoon Damrey swept westward across
the Indochina Peninsula as a tropical storm
Table 1. Three flood events of Chiang Mai in 2005
The Ping River Basin is one of the eight sub-basins in Chao
Phraya Basin. It stretches from latitude 19.75ºN to 15.75ºN
and from longitude 98.10ºE to 100.20ºE, with a catchment
area of 34,453 km2 (Figure 2). It covers about 22% of the Chao
Phraya River Basin and contributes about 24% (9044×106 m3)
of the total average annual runoff. Terraced mountains mainly
characterize the topography of Ping River Basin. About 55.5%
of total basin area is in the elevation range of 500–1500 m.
4.
METHODOLOGY
The extent and severity of damage from flooding are usually
defined by water depth. Such an inundation analysis can be
carried out effectively and efficiently by using numerical
modeling tools on a GIS platform. This also provides a
framework for the decision-support system and facilitates
evaluation of alternatives for flood management.
Chiang Mai is located in the north of Thailand, about 720
kilometers from Bangkok at an elevation of 1,027 feet (310 m.)
above sea level. To the North it borders Myanmar’s Shan State
while to the South it connects with Sam Ngao district of Tak
province. Chiang Mai’s geography comprises mainly groves
and mountains with a broad plain in the middle of the region on
both sides of Ping River. The province covers an area of
20,107.057 square kilometers (12,566,910 rai), made up of
8,787,656 rai (69.92%) of forest, 1,611,283 rai (12.82%) of
agricultural land and 2,167,971 rai (17.25%) of residential and
other land.
For this work, the HEC-RAS version 3.1 was used to calculate
water surface profiles; ArcView GIS 3.3 was used for GIS data
processing. The HEC-GeoRAS 3.1 for ArcView GIS was used
to provide the interface between the systems. HEC-GeoRAS is
an ArcView GIS extension specifically designed to process
geospatial data for use with HEC-RAS. The extension allows
users to create an HEC-RAS import file containing geometric
attribute data from an existing geographical data and
complementary data sets. GeoRAS automates the extraction of
spatial parameters for HEC-RAS input, primarily the
three-dimensional (3D) stream network and the 3D
cross-section definition. Results exported from HEC-RAS are
also processed in Geo-RAS. The ArcView 3D Analyst
extension is required to use GeoRAS.
4.1
Hydrological Data Selection
The general procedure adopted for inundation modeling
consists basically of four steps:
i) GeoRAS pre-processing to generate a HEC-RAS
import file,
ii) Running of HEC-RAS to calculate water surface
profiles,
iii) Post-processing of HEC-RAS results, and
iv) Flood hazard mapping.
Figure 3 explains these procedures in flow diagram.
Figure 2. Location map of Ping River Basin
205
Figure 4. TIN generated from spot heights
4.3
Running GeoRAS
This step was to generate the flood depth maps for the year
2005 September. Since the lack of the instantaneous peak
discharge data, the daily average data for the period of 1983 to
2004 were considered for analysis. The highest values of
discharge of particular months were taken in as a representation
for the month. Boundary conditions for the upstream as well as
the downstream were chosen in terms of water levels
corresponding input discharges for the sixteen gauge stations.
Steady flow analysis was performed for the month of
September.
4.4
Figure 3.
4.2
Post-processing of HEC-RAS
Model result of inundation area was compared with the Landsat
(ETM+) taken during the flood event of 2005. Flood depth was
compared with the field surveyed data generated by the survey
on the floodplain during the field visit at the selected locations
as a part of model output verification.
1D floodplain analysis using HEC-RAS
GeoRAS Pre-processing and TIN Generation
The main purpose of this step was to generate the geometric
data which was the DEM for HEC-RAS input from the existing
topographic and bathymetric data. The topographic data
consisted of the spot heights of the flood plain taken from the
surveyed data, the WGS84 ellipsoidal heights of the flood plain,
bathymetric data from Ping River and the GPS surveyed data
collected along the river banks.
4.5
Hazard Mapping
Hazard mapping was prepared by using the population data, the
land use data, and the flood depth in the GIS-base environment.
Population data was generated from the number of population
in each village. Land use was classified into four classes and
internal weight was assigned. Area for hazard mapping was
selected in the communes affected by flood. In total, there were
949 villages and 14 districts with an area of 1583.53km2. Flood
depth was reclassified into low, medium, high and very high
and internal weight was given to each class. In the same way,
weighted population was reclassified into low, medium, high
and very high and internal weight was assigned respectively.
DEM was created form all these themes in Arc-View GIS
software. A land use map of the study area was classified from
the Landsat7 (ETM+) image of April 08, 2005. Manning’s
roughness coefficient values were derived from the land use
map for each pixel based on the land use classes.
Cross-sections were selected at 45 positions at the river after
the river centerline and the river banks delineation were
completed as part of the geometric data formulation. These
cross sections included the 16 gauge stations along the river.
Geometric correction in HEC-RAS incorporated checking the
Manning’s roughness coefficient values for each cross-section
and restraining twenty values per cross section as of the model
constraint and bank-line modifications.
Theme
Reclass
Population
The TIN model was generated from the spot heights acquired
from different sources in ArcView GIS which included:
1) GPS surveyed data collected along the two river banks,
accuracy 5 meters.
2) The spot heights of the flood plains taken from the surveyed
data in 2003, accuracy 2.5 meter, (Source: Land development
Department)
3) River bed cross section elevation data accuracy 15
centimeter, (Source: Department of Water Resource).
Low
Medium
High
Very High
Class 1
Land Use
Flood
Depth
Class 2
Class 3
Class 4
Low
Medium
High
Very High
Class Indicators
0-3,000
3,000-6,000
6,000-10,000
10,000-25,000
Forest cover + Water
body
Grasslands
Agricultural area
Built-up area
(0-0.2) meters
(0.2-0.5) meters
(0.5-1.0) meters
＞1.0 meters
Table 2. Reclassification and internal weight
206
Internal
Weight
1
2
3
4
0
2
3
5
1
2
3
4
AHP (Analytic Hierarchy Process) was used for assigning the
final weight into each factor. The final flood hazard map after
calculation was reclassified into low, medium, high and very
high. The criteria for the use of AHP are:
¾
Flood depth is 2 times as important as Population
¾
Population is 3 times as important as Land use
¾
Flood depth is 4 times as important as Land use
Population
Flood depth
Land use
Population
1
2
1/3=0.33
Flood depth
1/2=0.5
1
1/4=0.25
Land use
3
4
1
Figure 6. Landsat image (left) compared with flood extent from
HEC-RAS (right)
Table 3. Criteria for AHP analysis
5.
5.1
RESULTS
Flood Extent and Depth (HEC-RAS)
As for the Figure 5 it is noticed that the depth varies from
0~1.68 m in the flood plain and on the river. The total flooded
area is 1,579 km2. Maximum area was inundated from a flood
depth of 1.09 – 1.68 m. That indicates how large the year 2005
flood was. Flood at 1 meter depth or more is probably
sufficient to cause damage to any built-up area if it stays for
some time.
Figure 7. Flood depth verification points (left)
Comparison of HEC depth with surveyed depth (right)
5.3
Flood Hazard Mapping
The final hazard map was calculated using equation from AHP
equations. It is seen from the Figure 8 that the total area under
hazard was 1,579 km2. Out of this area, 274 km2 was low
hazard, 410 km2 was medium hazard while 555 km2 was high
hazard and 338 km2 was under very high hazard category.
(a)
(b)
(c)
(d)
Figure 5. Flood depth of the study area from HEC-RAS
5.2
Comparison with the Base Data
The model result was verified with the Landsat (ETM+) image
(Source: Land Development Department). It is seen that the
model result reasonably matches with the Lamdsat.
Verification of depth was carried out with the survey at 30
points in the flood plains during the field visit by interviewing
people. The result matched quite closely in Mae Rim, Sansay
and Muang districts but a little differed in Jom Thong district.
Another area of uncertainty was the high values of depth at the
boundaries of the flood plain which should be verified since
HEC-RAS may act as a wall to the water flowing across the
boundary and thus computing a high value particularly in the
southern of Salaphi and Sankamphang districts. This can be
overcome by extending the area of modeling and that needs
more survey and elevation data.
Figure 8. Food hazard maps
(a) Flood hazard map;
(b) Affected school under each flood hazard category;
(c) Affected hospital under each flood hazard category;
(d) Affected factory under each flood hazard category
From the above results, we can get the conclusions that schools,
hospitals, factories affected by the flood of 2005 were
calculated by overlaying them with the final hazard map
(Figure 8(a)). And there are 590 schools, 142 hospitals and 451
factories were affected by the flood of 2005.
6.
207
CONCLUSION AND RECOMMENDATIONS
Flood hazard mapping is an important component for
appropriate land use planning in flood plain areas. It creates
easily-read, rapidly-accessible charts and maps which
facilitates the administrators and planners to identify areas of
risk and prioritize their mitigation or response efforts. This
paper presents an efficient methodology to accurately delineate
the flood-hazard areas in the Chiang Mai province; Northern
Thailand in a GIS based analysis. The study has used one of the
multi-criteria
decision-making
techniques,
Analytical
Hierarchical Process (AHP) which provides a systematic
approach for assessing and integrating the impact of various
factors, involving the levels of dependent and independent,
qualitative and quantitative information. Furthermore, valuable
advantage of HEC-RAS software is that it is readily available
for free download at HEC-RAS website.
Environment, University of Queensland, Brisbane, 15-17th
February 1994.
Chia Aik Song (2001). Flood extent in the lower Mekong basin
evaluated using spot quicklook mosaics, Paper presented at the
22nd Asian Conference on Remote Sensing, 5 – 9 November
2001, Singapore.
Coburn, (1994). Vulnerability and Risk Assessment, Disaster
Management Training Programme module, United Nations
Development Programme, Cambridge, United Kingdom (pp.
67-68 )
Zhou Chenghu. (1993). The Study of Flood and Waterlog
Disaster by Remote Sensing Monitoring, Geography Research,
1993, 12(2).
The method presented proves very applicable to adopt for
further study as can be seen from the verification step that the
model outputs either the inundation area or depth were
reasonably close to the satellite image and surveyed data
respectively. The summarized results can be concluded as
follows:
1) Inundation area from HEC-RAS shows that the flood depth
varies from 0~1.68 m in the flood plains and on the river. The
total flooded area is 1,579 km2.
2) Flood inundation area by and large closely agrees with the
Landsat image of the study area.
3) Flood depth from HEC-RAS is quite close to the surveyed
points.
4) Flood hazard analysis shows that maximum area under high
hazard category.
5) Most of the areas affected by the flood were agricultural
areas.
6) Most number of affected schools, hospitals, and factories are
under very high flood hazard.
7) This approach can be extended to the other parts of the
basin.
Cooperative Research Centre for Catchment Hydrology, (2006).
Series on model choice: General approaches to modelling and
practical
issues
of
model
choice,
from
http://www.toolkit.net.au/pdfs/MC-1.pdf
Coroza, O., Evans, D. and Bishop, I. (1997). ‘Enhancing runoff
modelling with GIS’, Landscape and Urban Planning, vol.38,
no.1-2, p13-23.
Dutta, D. and Tingsanchali, T. (2003). Development of Loss
Functions for Urban Flood LossAnalysis in Bangkok,
Proceeding of the 2nd International Symposium on New
Technologies for Urban Safety of Mega Cities in Asia, ICUS,
University of Tokyo, pp. 229-238.
Dutta, D., Herath, S. and Musiake, K. (2003). A Mathematical
Model for Flood Loss Estimation, Journal of Hydrology
Elsevier Science, Volume 277, pp. 24-49.
Fedra, K., Winkelbauer, L. and Pantulu, V.R. (1991). An
Application in the Lower Mekong Basin, RR-91-19.
International Institute for Applied Systems Analysis. A-236l
Laxenburg,
Austria.
169p,
from
http://www.ess.co.at/EIA/rr04.html
The results described in this study should provide helpful
information about flood hazard management and should be
useful in assigning priority for the development of very high
risk areas for flood control planning, and the construction and
development of flood countermeasures.
ACKNOWLEDGEMENTS
In addition, it is evident that flood hazard mapping at
pre-feasibility level could be carried out using secondary
information from maps, satellite images, and published and
unpublished documents.
The author would like to acknowledge the Chinese Academy of
Surveying and Mapping and Asian Institute of Technology for
providing a small project that enabled us to undertake this
study.
Finally, this type of flood hazard map in digital form can be
used as a database to be shared among the various government
and non-government agencies responsible for the construction
and development of flood defense.
REFERENCE
Alam, J. (2003). Two Dimensional Urban Flood Modelling for
Real Time Flood Forecasting for Dhaka City Bangladesh, AIT
Thesis No. WM 02-06, Asian Institute of Technology.
Alkema, D. (2003). Flood Risk Assessment for EIA; an
Example of a Motorway Near Trento, Italy, Studi Trentini Di
Scienze Naturali, Acta Geologica, V.78, pp 147-153.
Apelt, C.J. (1994). Physical and numerical hydraulic modelling;
past, present and future, International Conference on
Hydraulics in Civil Engineering: Hydraulics Working with the
208